Core Concepts
Enhancing synthetic tabular data utility through the DSF-GAN architecture with downstream feedback.
Stats
To enhance the utility of synthetic samples, we propose a novel architecture called the Down-Stream Feedback Generative Adversarial Network (DSF-GAN).
Our experiments demonstrate improved model performance when training on synthetic samples generated by DSF-GAN, compared to those generated by the same GAN architecture without feedback.
Quotes
"Many directions for future work are possible."
"This research is another stepping stone in enabling synthetic data’s safe and efficient use in machine-learning tasks."